GDScript 2.0: Casting to Zero when using property

[Ajrarn777]

Godot version

4.0.16

System information

Windows 10

Issue description

When using a property and casting the value from an INT to a FLOAT, produces very strange result.

See to step to replicate.

It seems when we assign a property to a float, the internal variable type is corrupted.

Steps to reproduce

#===============================================================================
var Value:int = 0 :
	get:
		return Value
	set(v):
		Value = v
		
#===============================================================================
func Test():
	var f:float 
	var n:int 
	
	Value = 7
	f = Value
	n = int( f )   # Here : f = 7, and n = 0 ??? <<<<<<<<<<<<<<<<<<<<<<<< BAD !!!

	f = 4
	n = int( f )   # Here : f = 4, and n = 4 GOOD

Minimal reproduction project

No response

[KoBeWi]

prints(f, int(f)) prints 7 0 🤔
The bug only occurs if you use a getter.

[TheSofox]

Seems connected to Proposal 844 godotengine/godot-proposals#844

[anvilfolk]

Digging a little into this one since I can still replicate it :)

It appears that there is no conversion happening between the int from the getter and the float f.

	f = Value
	print(typeof(f) == TYPE_INT)

if you ad this print statement, it will print true, which it shouldn't, because f is supposed to be a float.

Furthermore, in those conditions it appears that the constructor for the variant of type int might actually skip a few steps and forget to assign the value? I'm digging in a little more.

[anvilfolk]

Ok, so I got a fix in, but I'm not convinced it's completely thorough.

Essentially getters were not their return address types in the compiler, meaning that the compiler wouldn't know to create an OPCODE_ASSIGN_TYPED_BUILTIN instead of an OPCODE_ASSIGN.

Somehow I still believe that it's possible to fool the compiler into creating an OPCODE_ASSIGN rather then the typed variant, which will result in the same issue: the float variant containing an int.

Which also doesn't solve problem number 2, which is that calling int() on something strange like a float variable with an int inside results in a zero.